新增支持enhanced-rtmp h265 推流 (#2694)

src/Extension/Factory.cpp

@@ -211,7 +211,7 @@ static CodecId getVideoCodecIdByAmf(const AMFValue &val){
     return CodecInvalid;
 }
 
-Track::Ptr getTrackByCodecId(CodecId codecId, int sample_rate = 0, int channels = 0, int sample_bit = 0) {
+Track::Ptr Factory::getTrackByCodecId(CodecId codecId, int sample_rate, int channels, int sample_bit) {
     switch (codecId){
         case CodecH264 : return std::make_shared<H264Track>();
         case CodecH265 : return std::make_shared<H265Track>();

src/Extension/Factory.h

@@ -21,6 +21,16 @@ namespace mediakit{
 
 class Factory {
 public:
+
+    /**
+     * 根据codec_id 获取track
+     * @param codecId 编码id
+     * @param sample_rate 采样率，视频固定为90000
+     * @param channels 音频通道数
+     * @param sample_bit 音频采样位数
+     */
+    static Track::Ptr getTrackByCodecId(CodecId codecId, int sample_rate = 0, int channels = 0, int sample_bit = 0);
+
     ////////////////////////////////rtsp相关//////////////////////////////////
     /**
      * 根据sdp生成Track对象

src/Extension/H264Rtmp.cpp

@@ -164,12 +164,8 @@ bool H264RtmpEncoder::inputFrame(const Frame::Ptr &frame) {
         //not config
         _rtmp_packet->buffer[1] = true;
         int32_t cts = pts - dts;
-        if (cts < 0) {
-            cts = 0;
-        }
         //cts
         set_be24(&_rtmp_packet->buffer[2], cts);
-
         _rtmp_packet->time_stamp = dts;
         _rtmp_packet->body_size = _rtmp_packet->buffer.size();
         _rtmp_packet->chunk_id = CHUNK_VIDEO;

src/Extension/H265Rtmp.cpp

@@ -30,25 +30,8 @@ H265Frame::Ptr H265RtmpDecoder::obtainFrame() {
 }
 
 #ifdef ENABLE_MP4
-/**
- * 返回不带0x00 00 00 01头的sps
- * @return
- */
-static bool getH265ConfigFrame(const RtmpPacket &thiz,string &frame) {
-    if (thiz.getMediaType() != FLV_CODEC_H265) {
-        return false;
-    }
-    if (!thiz.isCfgFrame()) {
-        return false;
-    }
-    if (thiz.buffer.size() < 6) {
-        WarnL << "bad H265 cfg!";
-        return false;
-    }
-
-    auto extra = thiz.buffer.data() + 5;
-    auto bytes = thiz.buffer.size() - 5;
 
+static bool decode_HEVCDecoderConfigurationRecord(uint8_t *extra, size_t bytes, string &frame) {
     struct mpeg4_hevc_t hevc;
     memset(&hevc, 0, sizeof(hevc));
     if (mpeg4_hevc_decoder_configuration_record_load((uint8_t *)extra, bytes, &hevc) > 0) {
@@ -62,9 +45,88 @@ static bool getH265ConfigFrame(const RtmpPacket &thiz,string &frame) {
     }
     return false;
 }
+
+/**
+ * 返回不带0x00 00 00 01头的sps
+ */
+static bool getH265ConfigFrame(const RtmpPacket &thiz, string &frame) {
+    if (thiz.getMediaType() != FLV_CODEC_H265) {
+        return false;
+    }
+    if (!thiz.isCfgFrame()) {
+        return false;
+    }
+    if (thiz.buffer.size() < 6) {
+        WarnL << "bad H265 cfg!";
+        return false;
+    }
+    return decode_HEVCDecoderConfigurationRecord((uint8_t *)thiz.buffer.data() + 5, thiz.buffer.size() - 5, frame);
+}
 #endif
 
 void H265RtmpDecoder::inputRtmp(const RtmpPacket::Ptr &pkt) {
+    if (_info.codec == CodecInvalid) {
+        // 先判断是否为增强型rtmp
+        parseVideoRtmpPacket((uint8_t *)pkt->data(), pkt->size(), &_info);
+    }
+
+    if (_info.is_enhanced) {
+        // 增强型rtmp
+        parseVideoRtmpPacket((uint8_t *)pkt->data(), pkt->size(), &_info);
+        if (!_info.is_enhanced || _info.codec != CodecH265) {
+            throw std::invalid_argument("Invalid enhanced-rtmp hevc packet!");
+        }
+        auto data = (uint8_t *)pkt->data() + 5;
+        auto size = pkt->size() - 5;
+        switch (_info.video.pkt_type) {
+            case RtmpPacketType::PacketTypeSequenceStart: {
+#ifdef ENABLE_MP4
+                string config;
+                if (decode_HEVCDecoderConfigurationRecord(data, size, config)) {
+                    onGetH265(config.data(), config.size(), pkt->time_stamp, pkt->time_stamp);
+                }
+#else
+                WarnL << "请开启MP4相关功能并使能\"ENABLE_MP4\",否则对H265-RTMP支持不完善";
+#endif
+                break;
+            }
+
+            case RtmpPacketType::PacketTypeCodedFramesX:
+            case RtmpPacketType::PacketTypeCodedFrames: {
+                auto pts = pkt->time_stamp;
+                if (RtmpPacketType::PacketTypeCodedFrames == _info.video.pkt_type) {
+                    // SI24 = [CompositionTime Offset]
+                    CHECK(size > 7);
+                    int32_t cts = (((data[0] << 16) | (data[1] << 8) | (data[2])) + 0xff800000) ^ 0xff800000;
+                    pts += cts;
+                    data += 3;
+                    size -= 3;
+                }
+                splitFrame(data, size, pkt->time_stamp, pts);
+                break;
+            }
+
+            case RtmpPacketType::PacketTypeMetadata: {
+                // The body does not contain video data. The body is an AMF encoded metadata.
+                // The metadata will be represented by a series of [name, value] pairs.
+                // For now the only defined [name, value] pair is [“colorInfo”, Object]
+                // See Metadata Frame section for more details of this object.
+                //
+                // For a deeper understanding of the encoding please see description
+                // of SCRIPTDATA and SSCRIPTDATAVALUE in the FLV file spec.
+                // DATA = [“colorInfo”, Object]
+                break;
+            }
+            case RtmpPacketType::PacketTypeSequenceEnd: {
+                // signals end of sequence
+                break;
+            }
+            default: break;
+        }
+        return;
+    }
+
+    // 国内扩展(12) H265 rtmp
     if (pkt->isCfgFrame()) {
 #ifdef ENABLE_MP4
         string config;
@@ -78,41 +140,42 @@ void H265RtmpDecoder::inputRtmp(const RtmpPacket::Ptr &pkt) {
     }
 
     if (pkt->buffer.size() > 9) {
-        auto total_len = pkt->buffer.size();
-        size_t offset = 5;
         uint8_t *cts_ptr = (uint8_t *)(pkt->buffer.data() + 2);
         int32_t cts = (((cts_ptr[0] << 16) | (cts_ptr[1] << 8) | (cts_ptr[2])) + 0xff800000) ^ 0xff800000;
         auto pts = pkt->time_stamp + cts;
-        while (offset + 4 < total_len) {
+        splitFrame((uint8_t *)pkt->data() + 5, pkt->size() - 5, pkt->time_stamp, pts);
-            uint32_t frame_len;
-            memcpy(&frame_len, pkt->buffer.data() + offset, 4);
-            frame_len = ntohl(frame_len);
-            offset += 4;
-            if (frame_len + offset > total_len) {
-                break;
-            }
-            onGetH265(pkt->buffer.data() + offset, frame_len, pkt->time_stamp, pts);
-            offset += frame_len;
-        }
     }
 }
 
-inline void H265RtmpDecoder::onGetH265(const char* pcData, size_t iLen, uint32_t dts,uint32_t pts) {
+void H265RtmpDecoder::splitFrame(const uint8_t *data, size_t size, uint32_t dts, uint32_t pts) {
-    if(iLen == 0){
+    auto end = data + size;
+    while (data + 4 < end) {
+        uint32_t frame_len = load_be32(data);
+        data += 4;
+        if (data + frame_len > end) {
+            break;
+        }
+        onGetH265((const char *)data, frame_len, dts, pts);
+        data += frame_len;
+    }
+}
+
+inline void H265RtmpDecoder::onGetH265(const char *data, size_t size, uint32_t dts, uint32_t pts) {
+    if (size == 0) {
         return;
     }
 #if 1
     _h265frame->_dts = dts;
     _h265frame->_pts = pts;
     _h265frame->_buffer.assign("\x00\x00\x00\x01", 4); // 添加265头
-    _h265frame->_buffer.append(pcData, iLen);
+    _h265frame->_buffer.append(data, size);
 
     // 写入环形缓存
     RtmpCodec::inputFrame(_h265frame);
     _h265frame = obtainFrame();
 #else
     // 防止内存拷贝，这样产生的265帧不会有0x00 00 01头
-    auto frame = std::make_shared<H265FrameNoCacheAble>((char *)pcData,iLen,dts,pts,0);
+    auto frame = std::make_shared<H265FrameNoCacheAble>((char *)data, size, dts, pts, 0);
     RtmpCodec::inputFrame(frame);
 #endif
 }
@@ -185,9 +248,6 @@ bool H265RtmpEncoder::inputFrame(const Frame::Ptr &frame) {
         // not config
         _rtmp_packet->buffer[1] = true;
         int32_t cts = pts - dts;
-        if (cts < 0) {
-            cts = 0;
-        }
         // cts
         set_be24(&_rtmp_packet->buffer[2], cts);
 
@@ -207,18 +267,16 @@ void H265RtmpEncoder::makeVideoConfigPkt() {
     int8_t flags = FLV_CODEC_H265;
     flags |= (FLV_KEY_FRAME << 4);
     bool is_config = true;
-    auto rtmpPkt = RtmpPacket::create();
+    auto pkt = RtmpPacket::create();
     // header
-    rtmpPkt->buffer.push_back(flags);
+    pkt->buffer.push_back(flags);
-    rtmpPkt->buffer.push_back(!is_config);
+    pkt->buffer.push_back(!is_config);
     // cts
-    rtmpPkt->buffer.append("\x0\x0\x0", 3);
+    pkt->buffer.append("\x0\x0\x0", 3);
 
     struct mpeg4_hevc_t hevc;
     memset(&hevc, 0, sizeof(hevc));
-    string vps_sps_pps = string("\x00\x00\x00\x01", 4) + _vps +
+    string vps_sps_pps = string("\x00\x00\x00\x01", 4) + _vps + string("\x00\x00\x00\x01", 4) + _sps + string("\x00\x00\x00\x01", 4) + _pps;
-                         string("\x00\x00\x00\x01", 4) + _sps +
-                         string("\x00\x00\x00\x01", 4) + _pps;
     h265_annexbtomp4(&hevc, vps_sps_pps.data(), (int)vps_sps_pps.size(), NULL, 0, NULL, NULL);
     uint8_t extra_data[1024];
     int extra_data_size = mpeg4_hevc_decoder_configuration_record_save(&hevc, extra_data, sizeof(extra_data));
@@ -227,13 +285,13 @@ void H265RtmpEncoder::makeVideoConfigPkt() {
         return;
     }
     // HEVCDecoderConfigurationRecord
-    rtmpPkt->buffer.append((char *)extra_data, extra_data_size);
+    pkt->buffer.append((char *)extra_data, extra_data_size);
-    rtmpPkt->body_size = rtmpPkt->buffer.size();
+    pkt->body_size = pkt->buffer.size();
-    rtmpPkt->chunk_id = CHUNK_VIDEO;
+    pkt->chunk_id = CHUNK_VIDEO;
-    rtmpPkt->stream_index = STREAM_MEDIA;
+    pkt->stream_index = STREAM_MEDIA;
-    rtmpPkt->time_stamp = 0;
+    pkt->time_stamp = 0;
-    rtmpPkt->type_id = MSG_VIDEO;
+    pkt->type_id = MSG_VIDEO;
-    RtmpCodec::inputRtmp(rtmpPkt);
+    RtmpCodec::inputRtmp(pkt);
 #else
     WarnL << "请开启MP4相关功能并使能\"ENABLE_MP4\",否则对H265-RTMP支持不完善";
 #endif

src/Extension/H265Rtmp.h

@@ -25,7 +25,7 @@ public:
     using Ptr = std::shared_ptr<H265RtmpDecoder>;
 
     H265RtmpDecoder();
-    ~H265RtmpDecoder() {}
+    ~H265RtmpDecoder() = default;
 
     /**
      * 输入265 Rtmp包
@@ -33,15 +33,16 @@ public:
      */
     void inputRtmp(const RtmpPacket::Ptr &rtmp) override;
 
-    CodecId getCodecId() const override{
+    CodecId getCodecId() const override { return CodecH265; }
-        return CodecH265;
-    }
 
 protected:
-    void onGetH265(const char *pcData, size_t iLen, uint32_t dts,uint32_t pts);
     H265Frame::Ptr obtainFrame();
 
+    void onGetH265(const char *data, size_t size, uint32_t dts, uint32_t pts);
+    void splitFrame(const uint8_t *data, size_t size, uint32_t dts, uint32_t pts);
+
 protected:
+    RtmpPacketInfo _info;
     H265Frame::Ptr _h265frame;
 };
 

src/Rtmp/Rtmp.cpp

@@ -258,6 +258,55 @@ void RtmpHandshake::random_generate(char *bytes, int size)
     }
 }
 
+CodecId parseVideoRtmpPacket(const uint8_t *data, size_t size, RtmpPacketInfo *info) {
+    RtmpPacketInfo save;
+    info = info ? info : &save;
+    info->codec = CodecInvalid;
+
+    CHECK(size > 0);
+    if (data[0] >> 7 == 1) {
+        // IsExHeader == 1
+        CHECK(size >= 5, "Invalid rtmp buffer size: ", size);
+        info->is_enhanced = true;
+        info->video.frame_type = (RtmpFrameType)((data[0] >> 4) & 0x07);
+        info->video.pkt_type = (RtmpPacketType)(data[0] & 0x0f);
+        if (memcmp(data + 1, "av01", 4) == 0) {
+            // AV1
+            info->codec = CodecAV1;
+        } else if (memcmp(data + 1, "vp09", 4) == 0) {
+            // VP9
+            info->codec = CodecVP9;
+        } else if (memcmp(data + 1, "hvc1", 4) == 0) {
+            // HEVC(H265)
+            info->codec = CodecH265;
+        } else {
+            WarnL << "Rtmp video codec not supported: " << std::string((char *)data + 1, 4);
+        }
+    } else {
+        // IsExHeader == 0
+        info->is_enhanced = false;
+        info->video.frame_type = (RtmpFrameType)(data[0] >> 4);
+        info->video.rtmp_codec = (RtmpVideoCodec)(data[0] & 0x0f);
+
+        switch (info->video.rtmp_codec) {
+            case RtmpVideoCodec::h264: {
+                CHECK(size >= 1, "Invalid rtmp buffer size: ", size);
+                info->codec = CodecH264;
+                info->video.h264_pkt_type = (RtmpH264PacketType)data[1];
+                break;
+            }
+            case RtmpVideoCodec::h265: {
+                CHECK(size >= 1, "Invalid rtmp buffer size: ", size);
+                info->codec = CodecH265;
+                info->video.h264_pkt_type = (RtmpH264PacketType)data[1];
+                break;
+            }
+            default: WarnL << "Rtmp video codec not supported: " << (int)info->video.rtmp_codec; break;
+        }
+    }
+    return info->codec;
+}
+
 }//namespace mediakit
 
 namespace toolkit {

src/Rtmp/Rtmp.h

@@ -269,5 +269,72 @@ private:
 //根据音频track获取flags
 uint8_t getAudioRtmpFlags(const Track::Ptr &track);
 
+enum class RtmpFrameType : uint8_t {
+    reserved = 0,
+    key_frame = 1, // key frame (for AVC, a seekable frame)
+    inter_frame = 2, // inter frame (for AVC, a non-seekable frame)
+    disposable_inter_frame = 3, // disposable inter frame (H.263 only)
+    generated_key_frame = 4, // generated key frame (reserved for server use only)
+    video_info_frame = 5, // video info/command frame
+};
+
+enum class RtmpVideoCodec : uint8_t {
+    h263 = 2, // Sorenson H.263
+    screen_video = 3, // Screen video
+    vp6 = 4, // On2 VP6
+    vp6_alpha = 5, // On2 VP6 with alpha channel
+    screen_video2 = 6, // Screen video version 2
+    h264 = 7, // avc
+    h265 = 12, // 国内扩展
+};
+
+enum class RtmpH264PacketType : uint8_t {
+    h264_config_header = 0, // AVC sequence header(sps/pps)
+    h264_nalu = 1, // AVC NALU
+    h264_end_seq = 2, // AVC end of sequence (lower level NALU sequence ender is not REQUIRED or supported)
+};
+
+enum class RtmpPacketType : uint8_t {
+    PacketTypeSequenceStart = 0,
+    PacketTypeCodedFrames = 1,
+    PacketTypeSequenceEnd = 2,
+
+    // CompositionTime Offset is implied to equal zero. This is
+    // an optimization to save putting SI24 composition time value of zero on
+    // the wire. See pseudo code below in the VideoTagBody section
+    PacketTypeCodedFramesX = 3,
+
+    // VideoTagBody does not contain video data. VideoTagBody
+    // instead contains an AMF encoded metadata. See Metadata Frame
+    // section for an illustration of its usage. As an example, the metadata
+    // can be HDR information. This is a good way to signal HDR
+    // information. This also opens up future ways to express additional
+    // metadata that is meant for the next video sequence.
+    //
+    // note: presence of PacketTypeMetadata means that FrameType
+    // flags at the top of this table should be ignored
+    PacketTypeMetadata = 4,
+
+    // Carriage of bitstream in MPEG-2 TS format
+    // note: PacketTypeSequenceStart and PacketTypeMPEG2TSSequenceStart
+    // are mutually exclusive
+    PacketTypeMPEG2TSSequenceStart = 5,
+};
+
+struct RtmpPacketInfo {
+    CodecId codec = CodecInvalid;
+    bool is_enhanced;
+    union {
+        struct {
+            RtmpFrameType frame_type;
+            RtmpVideoCodec rtmp_codec;
+            RtmpPacketType pkt_type;
+            RtmpH264PacketType h264_pkt_type;
+        } video;
+    };
+};
+// https://github.com/veovera/enhanced-rtmp/blob/main/enhanced-rtmp-v1.pdf
+CodecId parseVideoRtmpPacket(const uint8_t *data, size_t size, RtmpPacketInfo *info = nullptr);
+
 }//namespace mediakit
 #endif//__rtmp_h

src/Rtmp/RtmpDemuxer.cpp

@@ -75,14 +75,20 @@ bool RtmpDemuxer::loadMetaData(const AMFValue &val){
             }
             if (key == "videodatarate") {
                 videodatarate = val.as_integer();
+                _videodatarate = videodatarate * 1024;
                 return;
             }
         });
         if (videocodecid) {
             // 有视频
             ret = true;
+            if (videocodecid->type() == AMF_NUMBER && videocodecid->as_integer() == (int)RtmpVideoCodec::h264) {
+                // https://github.com/veovera/enhanced-rtmp/issues/8
+                _complete_delay = true;
+            } else {
                 makeVideoTrack(*videocodecid, videodatarate * 1024);
             }
+        }
         if (audiocodecid) {
             // 有音频
             ret = true;
@@ -92,7 +98,7 @@ bool RtmpDemuxer::loadMetaData(const AMFValue &val){
         WarnL << ex.what();
     }
 
-    if (ret) {
+    if (ret && !_complete_delay) {
         // metadata中存在track相关的描述，那么我们根据metadata判断有多少个track
         addTrackCompleted();
     }
@@ -108,8 +114,14 @@ void RtmpDemuxer::inputRtmp(const RtmpPacket::Ptr &pkt) {
         case MSG_VIDEO: {
             if (!_try_get_video_track) {
                 _try_get_video_track = true;
-                auto codec = AMFValue(pkt->getMediaType());
+                RtmpPacketInfo info;
-                makeVideoTrack(codec, 0);
+                auto codec_id = parseVideoRtmpPacket((uint8_t *)pkt->data(), pkt->size(), &info);
+                if (codec_id != CodecInvalid) {
+                    makeVideoTrack(Factory::getTrackByCodecId(codec_id), _videodatarate);
+                    if (_complete_delay) {
+                        addTrackCompleted();
+                    }
+                }
             }
             if (_video_rtmp_decoder) {
                 _video_rtmp_decoder->inputRtmp(pkt);
@@ -133,11 +145,15 @@ void RtmpDemuxer::inputRtmp(const RtmpPacket::Ptr &pkt) {
 }
 
 void RtmpDemuxer::makeVideoTrack(const AMFValue &videoCodec, int bit_rate) {
+    makeVideoTrack(Factory::getVideoTrackByAmf(videoCodec), bit_rate);
+}
+
+void RtmpDemuxer::makeVideoTrack(const Track::Ptr &track, int bit_rate) {
     if (_video_rtmp_decoder) {
         return;
     }
     // 生成Track对象
-    _video_track = dynamic_pointer_cast<VideoTrack>(Factory::getVideoTrackByAmf(videoCodec));
+    _video_track = dynamic_pointer_cast<VideoTrack>(track);
     if (!_video_track) {
         return;
     }

src/Rtmp/RtmpDemuxer.h

@@ -45,12 +45,15 @@ public:
 
 private:
     void makeVideoTrack(const AMFValue &val, int bit_rate);
+    void makeVideoTrack(const Track::Ptr &val, int bit_rate);
     void makeAudioTrack(const AMFValue &val, int sample_rate, int channels, int sample_bit, int bit_rate);
 
 private:
     bool _try_get_video_track = false;
     bool _try_get_audio_track = false;
+    bool _complete_delay = false;
     float _duration = 0;
+    int _videodatarate = 0;
     AudioTrack::Ptr _audio_track;
     VideoTrack::Ptr _video_track;
     RtmpCodec::Ptr _audio_rtmp_decoder;

src/Rtmp/RtmpPlayer.cpp

@@ -191,6 +191,13 @@ void RtmpPlayer::send_connect() {
     obj.set("audioCodecs", (double) (0x0400));
     //只支持H264
     obj.set("videoCodecs", (double) (0x0080));
+
+    AMFValue fourCcList(AMF_STRICT_ARRAY);
+    fourCcList.add("av01");
+    fourCcList.add("vp09");
+    fourCcList.add("hvc1");
+    obj.set("fourCcList", fourCcList);
+
     sendInvoke("connect", obj);
     addOnResultCB([this](AMFDecoder &dec) {
         //TraceL << "connect result";

src/Rtmp/RtmpPusher.cpp

@@ -135,6 +135,13 @@ void RtmpPusher::send_connect() {
     obj.set("type", "nonprivate");
     obj.set("tcUrl", _tc_url);
     obj.set("swfUrl", _tc_url);
+
+    AMFValue fourCcList(AMF_STRICT_ARRAY);
+    fourCcList.add("av01");
+    fourCcList.add("vp09");
+    fourCcList.add("hvc1");
+    obj.set("fourCcList", fourCcList);
+
     sendInvoke("connect", obj);
     addOnResultCB([this](AMFDecoder &dec) {
         //TraceL << "connect result";